基于组合估计量的域样本量分配方法研究

The Study of Domain Sample Size Allocation based on Composite Estimation

域估计问题是抽样调查的研究热点问题，若为计划域且总体样本量已知时，域的样本量分配是十分重要的。域样本量的分配方法包含比例分配、最优分配、内曼分配、等量分配、Power分配和Longford分配，这些方法基于较好的抽样设计，并没有结合域的估计方法，基于域的组合估计量并结合Longford损失函数得到组合分配和限制组合分配方法。最后，使用CFPS2010数据对等量分配、比例分配、组合分配和限制组合分配进行比较研究，说明组合分配方法的效率最高，是较好的域样本量分配方法。

The domain estimation is research hotspots in survey sampling. When the domain is the planned domain and the total sample size is fixed. The sample size allocation of domain is very important. There are proportional allocation, optimum allocation, Neyman allocation, equal allocation, Power allocation and Longford allocation in sample size allocation of domain. But, these methods are based on better sampling design and don＇t use domain estimation methods. In this paper, we use composite estimation of domain and Longford loss function, then get the composite allocation and composite allocation with constraints. Finally, a comparative study of the equal allocation, proportional allocation, composite allocation and composite allocation with constraints using the data of Chinese family panel study in this paper, further illustrate the efficiency of the composite allocation is the highest, it is the better sample size allocation domain.